software and data

software:

wfmoments
fastDTWF
vilma
pyrho
defiNETti

data:

human recombination maps
neanderthal introgression calls
gene-level loss-of-function constraint

wfmoments

fast calculation of heterozygosity under spatial models

wfmoments can compute the equilibrium and dynamics of expected heterozygosity under arbitrary deme models. wfmoments will be described in a forthcoming paper.

fastDTWF

fast calculation of likelihoods under the Discrete-Time Wright-Fisher model

fastDTWF implements machinery for computing various likelihoods under the Discrete-Time Wright-Fisher model. This includes computing frequency spectra and transition mass functions. The machiney can handle quite general models incorporating mutation and selection. fastDTWF is described in this paper.

vilma

polygenic scores using variational inference on gwas summary statistics from multiple cohorts

vilma is a method for building polygenic scores and inferring effect size distributions (genetic archtecture) from gwas summary statistics and ld reference panels. vilma can be applied to one or two cohorts. vilma is described in this paper.

pyrho

fast inference of fine-scale recombination rates

pyrho is a method for inferring fine-scale recombination maps from sequencing data from unrelated individuals while taking demographic history into account. it is substantially faster than LDhat while attaining comparable or better accuracy. in our paper we use pyrho to infer fine-scale recombination maps for the 26 populations in phase 3 of the thousand genomes project for the two most recent builds of the human genome (hg19 and hg38). those maps are available here.

defiNETti

exchangeable neural networks

defiNETti is a wrapper for TensorFlow that automatically builds exchangeable neural networks (i.e., networks that enforce permutation-invariance) and trains them using ‘simulation-on-the-fly’ as described in our paper. named after bruno.

fine-scale human recombination maps

accurate, population-specific fine-scale recombination maps for 26 populations

we inferred fine-scale recombination maps for each of the 26 populations in phase 3 of the thousand genomes project using our method pyrho. the maps were inferred while taking demography into account and explain patterns of ld better than previous recombination maps. the method and recombination maps are described in our paper. these recombination maps are also available for use in simulations via the wonderful stdpopsim project.

neanderthal introgression calls

posterior probabilities of introgression in modern humans

we called tracts of neanderthal introgression in ceu, chb, and chs using our method, dical-admix. the method and an analysis of the tracts are presented in our paper.

gene-level loss-of-function constraint

estimates of the strength of selection against loss-of-function variants

we developed powerful machinery to combine information across genes, GeneBayes, to estimate the strength of selection acting against loss-of-function variants in almost every gene in the human genome. the method and an analysis of the constraint measures are presented in our paper.